Introduction The publication of Patricia Crone and Michael Cook''s controversial study Hagarism in 1977 unquestionably marks a watershed in the study of religious culture in the early medieval Near East, even if its significance has occasionally been underestimated by other specialists in this field. In particular, this relatively slim volume highlighted the potential importance of non-Islamic literature for knowledge of religious (and secular) history in the seventh and eighth centuries, a so-called dark age for which sources are often sparse and spotty. Perhaps more importantly, however, this study proposed a radical new model for understanding both the formation of the Islamic tradition and the general religious landscape of the early medieval Near East. Together with the contemporary works of John Wansbrough, Hagarism articulated an innovative reinterpretation of formative Islam as a faith intimately intertwined with the religious traditions of Mediterranean late antiquity and in need of extensive study in the context of this religiously complex and intercultural milieu. There are, it must be admitted, some considerable and undeniable flaws in Hagarism ''s reinterpretation of formative Islam, as even its most sympathetic readers have often acknowledged. Most significantly, Hagarism has been rightly criticized for its occasionally uncritical use of non-Islamic sources in reconstructing the origins of Islam. Wansbrough, for instance, asks rather pointedly of Crone and Cook''s reconstruction: "Can a vocabulary of motives be freely extrapolated from a discrete collection of literary stereotypes composed by alien and mostly hostile observers, and thereupon employed to describe, even interpret, not merely the overt behavior but also the intellectual and spiritual development of helpless and mostly innocent actors?" Undoubtedly, Wansbrough''s question is intended as rhetorical and meant to impugn the value of non-Islamic sources for understanding earliest Islam. Nonetheless, I think that the most honest and accurate answer to this question is in fact, possibly.

While such information perhaps cannot be freely extracted from these sources, when analyzed with some care they may potentially yield historically valuable information concerning the beginnings of Islam. The imperfections of Hagarism should not lead us to discount completely the important insights that both this study and its approach have to offer. While some scholars have somewhat unfairly dismissed Hagarism and its approach as either hopelessly colonialist or methodologically flawed, there is still much to gain from this seminal book. Wansbrough''s more considered rejection of Hagarism reflects his concern for the overwhelming and historically distorting impact of "salvation history," that is, theologized, sacred history, on both the Islamic and non-Islamic sources, and in light of this he essentially committed himself to an historical agnosticism regarding the origins of Islam. Yet such resignation is not our only option. Admittedly, both Wansbrough and Robert Hoyland after him have correctly noted that non-Muslim sources alone "cannot provide a complete and coherent account of the history of Early Islam," as was essentially proposed in Hagarism . But this recognition does not somehow make non-Islamic witnesses to the religious history of the seventh and eighth centuries any less valuable as a whole than the early Islamic sources, and on particular points they may possibly report more reliably than the Islamic tradition, as this study will argue. Almost all the documentary resources for understanding the formative period of Islam, including even the Qur''an, are highly problematic from a religious historian''s viewpoint: these sources are frequently overwhelmed and controlled by a master narrative of sacred history, as well as being influenced by the social, political, and theological concerns of the particular groups that produced them.

But such conditions do not present an altogether uncommon or impossible circumstance. There are ways of extracting historically credible data from such "contaminated" repositories. We must deploy methods capable of identifying different types of bias and excavating information from these sources, along the lines of those techniques used to reconstruct the historical Jesus from the highly theologized narratives of the Christian gospels. This endeavor will not yield, to be sure, history "wie es eigentlich gewesen," but this was always a hypermodern fantasy in any case. Instead, we will be able to reconstruct a narrative (or quite possibly several narratives) of Islamic origins that possesses a degree of probability derived from the particular methodological principles used to assess the relative reliability of various testimonies concerning the formation of Islam. Hagarism opened the door to this new approach, and in its wake we must critically assess the strikingly dissimilar descriptions of earliest Islam often found in the non-Islamic sources of the seventh and eighth centuries and in the more traditional Islamic accounts from the later eighth and ninth centuries. While a great deal of investigation still remains to be done along these lines, the past two decades have already seen some excellent work in this area, much of it inspired by the initial insights of Crone and Cook. Rather than pursuing one of the many new issues that undoubtedly await exploration, the present study will return to what was surely one of Hagarism ''s most startling revelations: its identification of widespread reports from seventh- and eighth-century writers that Muhammad was still alive and leading the Islamic community as his followers began their invasion of the Roman Near East.

This indication is strikingly at odds with the traditional account of Muhammad''s death before the Near Eastern conquest at Medina in 632, first recorded in the earliest Islamic biographies of the mid-eighth and ninth centuries. With so many unanswered questions still to pursue, one might rightly question the return to an issue raised now already over thirty years ago. There are, however, several reasons for doing so. In the first place, Crone and Cook merely note the existence of this discrepancy in the sources, gathering many of the most significant references together in an endnote. Instead of carefully evaluating the historical significance of these witnesses both individually and collectively, they conclude their list of references with only the remark: "The convergence is impressive." Indeed it is, but can we say something more than this? Might a critical analysis of the sources give us some sense of how much historical weight they can bear, both individually and collectively? Is it possible that, even if Muhammad did not in fact lead the Islamic conquest of Palestine, this tradition might reveal something about the nature of formative Islam? In all fairness, we are presently much better equipped to pose such questions, in large part due to the excellent work of Hoyland, most notably in his Seeing Islam as Others Saw It. Not only has Hoyland produced an outstanding catalogue of the many references to early Islam made in non-Islamic sources, but he takes the project that was begun in Hagarism an important step forward by proposing a basic methodology for evaluating the significance of these sources, as well as providing examples of its application. In essence, Hoyland proposes that we should ask three basic questions of each potential witness to assess its historical worth: What is the source of its observation(s) about early Islam? What is the character of the observation? And what is the subject of the observation? The first question rather straightforwardly asks us to consider the reliability of each author''s source: Was he himself an eyewitness to what he reports? Did she hear it from those who were eyewitnesses? Or is it merely hearsay or gossip? Clearly there is a descending scale of reliability as one moves down this list.

In addition, Hoyland suggests that we consider the nature of the observation itself: does the source report a "simple observation of fact," or does the information in question serve some sort of apologetic agenda or "totalizing explanation"? "Simple observations," Hoyland suggests, will likely have a much higher degree of historical veracity. Somewhat related to this is the third principle, which questions the nature of the matter that the non-Islamic source describes: Is it something that an outsider would likely have accurate knowledge of? That is, does the statement reflect something that would be readily observable by a non-Muslim, or even better, is it something that would have directly affected non-Muslims? In such cases, the witness of non-Muslim writers is more likely to transmit reliable information. When the same writers comment on aspects of Islamic belief and intracommunal life, however, we must adopt a more skeptical approach to their reports. These are sound principles for assessing the relative worth of the various non-Islamic witnesses to the earliest history of Islam, to which I would add one further: the criterion of multiple, independent attestation, one of the oldest and most fundamental principles of modern biblical criticism and particularly important for studies of the historical Jesus. As biblical scholars have long recognized, a higher degree of historical probability inheres in observations attested by several independent sources, since this pattern makes it highly unlikely that a particular writer has invented a given report. When a particular tradition from the non-Islamic sources meets all of these criteria, there is a significant probability that such a report reflects genuine information about the formative period of Islam. While it cannot be said with any certainty that these w.